Scott Alexander on AI labs should give independent evaluat…

Scott Alexander

Author and psychiatrist

Thank you!

AI Verified (Sep 12, 2026)

Source:

Scott Alexander (@slatestarcodex), X post, September 12, 2026 (UTC): “Thank you!” The post quote-tweets Dario Amodei’s proposal stating that Anthropic will provide third-party evaluators with permanent, employee-level access to its systems to verify safety practices, report incidents, and assess model alignment during training.

20d ago

Policy proposals and claims

AI labs should give independent evaluators employee-like access

votes For

Statement relation verification history AI Verified Report this

AI Verified The quote is directly attached to Scott Alexander’s quote-tweet of Dario Amodei’s proposal for independent evaluators with employee-like access. In that source context, “Thank you!” is an affirmative acknowledgment of the complete evaluator-access proposal, making relevance and a supportive stance determinable. · Hector Perez Arenas gpt-5 · 20d ago

Vote inference verification history AI Verified Report this

AI Verified The vote author is Scott Alexander, matching the quote author. His “Thank you!” quote-tweet positively acknowledges Dario Amodei’s employee-like independent-evaluator access proposal, so the recorded answer “for” is correct. · Hector Perez Arenas gpt-5 · 20d ago

Quote authenticity verification history

Report this

AI Verified The supplied X post is attributed to Scott Alexander and dated September 12, 2026 UTC; the accessible syndicated record contains the exact text “Thank you!” and shows it quote-tweets Dario Amodei’s evaluator-access proposal. · Hector Perez Arenas gpt-5 · 20d ago

replying to Scott Alexander

Other opinions from Scott Alexander

See all

Scott Alexander votes For

We should adopt AI 2040 Plan A

If so, is it as good as Plan A? If not, consider demanding that they do better.

AI Verified source (Jul 10, 2026)

added 27d ago

Scott Alexander votes For

The U.S. and China should establish Mutually Assured Compute Destruction

Plan A feels like our best hope to do something actually good.

AI Verified source (Jul 9, 2026)

added 2mo ago

Scott Alexander votes For

Pausing AI at human-genius level to solve alignment is safer than either racing to superintelligence or halting entirely

Then, in the mid-2030s, they pause at AIs around the level of top human geniuses. These AIs are close enough to current AIs that the same alignment techniques that mostly-sort-of-work for ours might mostly-sort-of-work for them too... The plan is to ...

more AI Verified source (Jul 9, 2026)

added 2mo ago